File Download
 
Links for fulltext
(May Require Subscription)
 
Supplementary

Article: Model selection in time series studies of influenza-associated mortality
  • Basic View
  • Metadata View
  • XML View
TitleModel selection in time series studies of influenza-associated mortality
 
AuthorsWang, XL1
Yang, L1
Chan, KP1
Chiu, SS1
Chan, KH2
Peiris, JSM1
Wong, CM1
 
KeywordsAcute respiratory tract disease
Bayes theorem
Generalized cross validation
Health hazard
Intermethod comparison
 
Issue Date2012
 
PublisherPublic Library of Science. The Journal's web site is located at http://www.plosone.org/home.action
 
CitationPlos One, 2012, v. 7 n. 6 [How to Cite?]
DOI: http://dx.doi.org/10.1371/journal.pone.0039423
 
AbstractBackground: Poisson regression modeling has been widely used to estimate influenza-associated disease burden, as it has the advantage of adjusting for multiple seasonal confounders. However, few studies have discussed how to judge the adequacy of confounding adjustment. This study aims to compare the performance of commonly adopted model selection criteria in terms of providing a reliable and valid estimate for the health impact of influenza. Methods: We assessed four model selection criteria: quasi Akaike information criterion (QAIC), quasi Bayesian information criterion (QBIC), partial autocorrelation functions of residuals (PACF), and generalized cross-validation (GCV), by separately applying them to select the Poisson model best fitted to the mortality datasets that were simulated under the different assumptions of seasonal confounding. The performance of these criteria was evaluated by the bias and root-mean-square error (RMSE) of estimates from the pre-determined coefficients of influenza proxy variable. These four criteria were subsequently applied to an empirical hospitalization dataset to confirm the findings of simulation study. Results: GCV consistently provided smaller biases and RMSEs for the influenza coefficient estimates than QAIC, QBIC and PACF, under the different simulation scenarios. Sensitivity analysis of different pre-determined influenza coefficients, study periods and lag weeks showed that GCV consistently outperformed the other criteria. Similar results were found in applying these selection criteria to estimate influenza-associated hospitalization. Conclusions: GCV criterion is recommended for selection of Poisson models to estimate influenza-associated mortality and morbidity burden with proper adjustment for confounding. These findings shall help standardize the Poisson modeling approach for influenza disease burden studies. © 2012 Wang et al.
 
Descriptioneid_2-s2.0-84862699414
 
ISSN1932-6203
2013 Impact Factor: 3.534
2013 SCImago Journal Rankings: 1.724
 
DOIhttp://dx.doi.org/10.1371/journal.pone.0039423
 
ReferencesReferences in Scopus
 
DC FieldValue
dc.contributor.authorWang, XL
 
dc.contributor.authorYang, L
 
dc.contributor.authorChan, KP
 
dc.contributor.authorChiu, SS
 
dc.contributor.authorChan, KH
 
dc.contributor.authorPeiris, JSM
 
dc.contributor.authorWong, CM
 
dc.date.accessioned2012-08-16T05:54:48Z
 
dc.date.available2012-08-16T05:54:48Z
 
dc.date.issued2012
 
dc.description.abstractBackground: Poisson regression modeling has been widely used to estimate influenza-associated disease burden, as it has the advantage of adjusting for multiple seasonal confounders. However, few studies have discussed how to judge the adequacy of confounding adjustment. This study aims to compare the performance of commonly adopted model selection criteria in terms of providing a reliable and valid estimate for the health impact of influenza. Methods: We assessed four model selection criteria: quasi Akaike information criterion (QAIC), quasi Bayesian information criterion (QBIC), partial autocorrelation functions of residuals (PACF), and generalized cross-validation (GCV), by separately applying them to select the Poisson model best fitted to the mortality datasets that were simulated under the different assumptions of seasonal confounding. The performance of these criteria was evaluated by the bias and root-mean-square error (RMSE) of estimates from the pre-determined coefficients of influenza proxy variable. These four criteria were subsequently applied to an empirical hospitalization dataset to confirm the findings of simulation study. Results: GCV consistently provided smaller biases and RMSEs for the influenza coefficient estimates than QAIC, QBIC and PACF, under the different simulation scenarios. Sensitivity analysis of different pre-determined influenza coefficients, study periods and lag weeks showed that GCV consistently outperformed the other criteria. Similar results were found in applying these selection criteria to estimate influenza-associated hospitalization. Conclusions: GCV criterion is recommended for selection of Poisson models to estimate influenza-associated mortality and morbidity burden with proper adjustment for confounding. These findings shall help standardize the Poisson modeling approach for influenza disease burden studies. © 2012 Wang et al.
 
dc.description.naturepublished_or_final_version
 
dc.descriptioneid_2-s2.0-84862699414
 
dc.identifier.citationPlos One, 2012, v. 7 n. 6 [How to Cite?]
DOI: http://dx.doi.org/10.1371/journal.pone.0039423
 
dc.identifier.doihttp://dx.doi.org/10.1371/journal.pone.0039423
 
dc.identifier.epagee39423
 
dc.identifier.hkuros202482
 
dc.identifier.issn1932-6203
2013 Impact Factor: 3.534
2013 SCImago Journal Rankings: 1.724
 
dc.identifier.issue6
 
dc.identifier.pmid22745751
 
dc.identifier.scopuseid_2-s2.0-84862699414
 
dc.identifier.spagee39423
 
dc.identifier.urihttp://hdl.handle.net/10722/159710
 
dc.identifier.volume7
 
dc.languageeng
 
dc.publisherPublic Library of Science. The Journal's web site is located at http://www.plosone.org/home.action
 
dc.publisher.placeUnited States
 
dc.relation.ispartofPLoS ONE
 
dc.relation.referencesReferences in Scopus
 
dc.rightsCreative Commons: Attribution 3.0 Hong Kong License
 
dc.subjectAcute respiratory tract disease
 
dc.subjectBayes theorem
 
dc.subjectGeneralized cross validation
 
dc.subjectHealth hazard
 
dc.subjectIntermethod comparison
 
dc.titleModel selection in time series studies of influenza-associated mortality
 
dc.typeArticle
 
<?xml encoding="utf-8" version="1.0"?>
<item><contributor.author>Wang, XL</contributor.author>
<contributor.author>Yang, L</contributor.author>
<contributor.author>Chan, KP</contributor.author>
<contributor.author>Chiu, SS</contributor.author>
<contributor.author>Chan, KH</contributor.author>
<contributor.author>Peiris, JSM</contributor.author>
<contributor.author>Wong, CM</contributor.author>
<date.accessioned>2012-08-16T05:54:48Z</date.accessioned>
<date.available>2012-08-16T05:54:48Z</date.available>
<date.issued>2012</date.issued>
<identifier.citation>Plos One, 2012, v. 7 n. 6</identifier.citation>
<identifier.issn>1932-6203</identifier.issn>
<identifier.uri>http://hdl.handle.net/10722/159710</identifier.uri>
<description>eid_2-s2.0-84862699414</description>
<description.abstract>Background: Poisson regression modeling has been widely used to estimate influenza-associated disease burden, as it has the advantage of adjusting for multiple seasonal confounders. However, few studies have discussed how to judge the adequacy of confounding adjustment. This study aims to compare the performance of commonly adopted model selection criteria in terms of providing a reliable and valid estimate for the health impact of influenza. Methods: We assessed four model selection criteria: quasi Akaike information criterion (QAIC), quasi Bayesian information criterion (QBIC), partial autocorrelation functions of residuals (PACF), and generalized cross-validation (GCV), by separately applying them to select the Poisson model best fitted to the mortality datasets that were simulated under the different assumptions of seasonal confounding. The performance of these criteria was evaluated by the bias and root-mean-square error (RMSE) of estimates from the pre-determined coefficients of influenza proxy variable. These four criteria were subsequently applied to an empirical hospitalization dataset to confirm the findings of simulation study. Results: GCV consistently provided smaller biases and RMSEs for the influenza coefficient estimates than QAIC, QBIC and PACF, under the different simulation scenarios. Sensitivity analysis of different pre-determined influenza coefficients, study periods and lag weeks showed that GCV consistently outperformed the other criteria. Similar results were found in applying these selection criteria to estimate influenza-associated hospitalization. Conclusions: GCV criterion is recommended for selection of Poisson models to estimate influenza-associated mortality and morbidity burden with proper adjustment for confounding. These findings shall help standardize the Poisson modeling approach for influenza disease burden studies. &#169; 2012 Wang et al.</description.abstract>
<language>eng</language>
<publisher>Public Library of Science. The Journal&apos;s web site is located at http://www.plosone.org/home.action</publisher>
<relation.ispartof>PLoS ONE</relation.ispartof>
<rights>Creative Commons: Attribution 3.0 Hong Kong License</rights>
<subject>Acute respiratory tract disease</subject>
<subject>Bayes theorem</subject>
<subject>Generalized cross validation</subject>
<subject>Health hazard</subject>
<subject>Intermethod comparison</subject>
<title>Model selection in time series studies of influenza-associated mortality</title>
<type>Article</type>
<description.nature>published_or_final_version</description.nature>
<identifier.doi>10.1371/journal.pone.0039423</identifier.doi>
<identifier.pmid>22745751</identifier.pmid>
<identifier.scopus>eid_2-s2.0-84862699414</identifier.scopus>
<identifier.hkuros>202482</identifier.hkuros>
<relation.references>http://www.scopus.com/mlt/select.url?eid=2-s2.0-84862699414&amp;selection=ref&amp;src=s&amp;origin=recordpage</relation.references>
<identifier.volume>7</identifier.volume>
<identifier.issue>6</identifier.issue>
<identifier.spage>e39423</identifier.spage>
<identifier.epage>e39423</identifier.epage>
<publisher.place>United States</publisher.place>
<bitstream.url>http://hub.hku.hk/bitstream/10722/159710/1/Content.pdf</bitstream.url>
</item>
Author Affiliations
  1. The University of Hong Kong
  2. Queen Mary Hospital Hong Kong